4D.3
Ensemble prediction of tropical cyclones
K. Puri, BMRC, Melbourne, Vic., Australia; and J. Barkmeijer, N. E. Davidson, and H. C. Weber
The last few years have seen significant advances in tropical cyclone (TC) track prediction using numerical weather prediction models. Several factors have contributed to these advances which include increases in model resolution, improvements in the parameterization of physical processes, improvements in data assimilation, and the implementation of numerics that are more accurate and stable. In spite of the impressive progress there is considerable variability in performance of models, and there can be large variation of tracks from one day to the next. Ensemble prediction, which has become an established part of operational global weather prediction at a number of centres, provides one possible means of addressing the problem of track uncertainty by providing estimates of the uncertainty of a deterministic track forecast.
A key problem in ensemble prediction is the generation of initial perturbations, and two types of dynamically-constrained perturbations have been proposed: one based on ‘bred' vectors is used at the National Centers for Environment Prediction (NCEP), while the other, based on a singular vector approach (SV) is used at ECMWF. To date the practical application of the ensemble prediction systems (EPSs) has been restricted to the extratropics. For the singular vector approach the principal reason for this restriction is that although the assumption of linear error growth simplifies the computation of fast-growing structures, it still requires a linearized version and adjoint of the forecast model. For the tropics, adiabatic linear models are not optimal and there is no obvious strategy to include physical processes in linear models because of their strong nonlinearities and the frequent dependence on logical conditionals. Recently, linear models including physical processes have been developed and at ECMWF the linearised physics package is being used to compute SVs. Another feature of the ECMWF system relevant to the tropics is the inclusion of a simulation of random errors associated with parameterized physics in the model which is referred to as stochastic physics. These developments provide a means to extend the application of ensemble prediction to the tropics.
Puri et al. (2001) have studied the feasibility of using initial perturbations generated using diabatic SVs for TC prediction in terms of spread in cyclone tracks and intensities and have shown that the procedure offers encouraging potential to predict the spread. A number of operational centres use TC bogus data as a means of generating an initial vortex. It is possible to perturb the initial vortex in a number of ways to provide an alternative method of generating initial perturbations for TC ensemble prediction. This latter approach has the advantage of being simple compared to the complexity of using diabatic singular vectors. In this paper the performance of the two very different techniques will be compared.
Reference
Puri, K., J. Barkmeijer, and T.N. Palmer, 2001: Ensemble prediction of tropical cyclones using targeted diabatic singular vectors. Quart. J. Roy. Meteor. Soc., 127, 709-731.
Session 4D, Tropical Cyclone Prediction II (Parallel with Sessions 4A, 4B, & 4C)
Tuesday, 30 April 2002, 8:30 AM-10:30 AM
Previous paper Next paper